|Publication number||US7631261 B2|
|Application number||US 10/660,747|
|Publication date||Dec 8, 2009|
|Priority date||Sep 12, 2002|
|Also published as||US20040169676|
|Publication number||10660747, 660747, US 7631261 B2, US 7631261B2, US-B2-7631261, US7631261 B2, US7631261B2|
|Inventors||Thomas D. Williams, Alfred A. Hough|
|Original Assignee||Inoue Technologies, LLC|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (16), Non-Patent Citations (5), Referenced by (3), Classifications (31), Legal Events (3)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This nonprovisional application claims the benefit of U.S. Provisional Application No. 60/410,335, filed Sep. 12, 2002.
The present invention generally relates to the field of imagery in multimedia, video (including interactive, immersive, Internet and networked streaming video), virtual reality, telepresence and television. More particularly, the present invention relates to a method for delivering continuous multiple visual telepresences to a large number of simultaneous users in real time.
It is clearly impractical to have many (perhaps tens of thousands) of cameras moving through a scene where each camera is physically controlled by an individual user to create a unique user-preferred view. Alternatively, a set of fixed cameras are placed in the environment and user-preferred views that move through the environment could be synthesized. Although computer systems of this era are powerful, the need to service so many users with unique viewpoint control is a very demanding requirement.
The present invention concerns the effective delivery of a visual virtual reality or a visual telepresence experience to many simultaneous users who are viewing an actual scene in real-time. Delivery of one or more of these views is accomplished with varying degrees of viewer control, as desired, depending on the application. The number of simultaneous users may range from a few people, to thousands or in the tens of thousands.
By way of overview, a plurality of fixed cameras are placed in the subject environment to be viewed, and virtual moving images are synthesized in accordance with the method of the present invention, to create unique visual telepresences. Referring now to
In a preferred embodiment, the method for efficiently delivering one or more visual virtual realities to a plurality of simultaneous users comprises the steps of:
The user viewing the video stream on a display screen sees a real-time viewer-centered moving image, for instance, as though the user was walking on a sidewalk in an actual town. The display presentation is real, in that all events that exist at the time of viewing are occurring substantially instantaneously, much as a high-speed web camera or television system displays remote actual events, for example.
Unlike a web camera or television system, the viewpoint displayed in the present invention moves in tandem with user movement. For example, if the application simulates a walking motion, the viewpoint displayed moves as though the viewer is walking. This is not the view from any existing camera.
In addition, the viewer is capable of changing the viewpoint's path as desired. For instance, a security application of the present invention may require the capability of cameras to move quickly, pan, and zoom and then to dwell, and then jump quickly to a distant viewpoint, whereas a shopping or tourist application would likely require less involved motion type attributes, such as moving slowly and continuously and/or dwelling.
Users may be electronically connected, locally or globally via a communications network, to navigate through the event to facilitate real-time viewing. This is facilitated by the use of multiple server computers each providing service to some number of users.
With these and other advantages and features of the invention that may become hereinafter apparent, the nature of the invention may be more clearly understood by reference to the following detailed description of the invention, the appended claims and to the several drawings attached herein.
The following terms, in alphabetical order, is presented for easier understanding of the detailed description of the present invention.
Continuous Video Stream
Digital video data that does not stop unless the viewer decides to terminate.
Happenings, happenstances, interesting visual on-goings.
Image (A Digital Image)
A regular array of point values, usually representing brightness and color values as samples of a picture. Sometimes an image is an array of values, still registered with the image frame, but representing quantities other than brightness or color, such as correspondences or image differences. An image may be one of a sequence, such as the digital representation of a fame of video.
Places in the environment where short sequence paths begin and end.
See “Synthetic Viewpoint”.
Predetermined Short Sequence, or “SS”.
A set of viewpoints, determined in advance, that starts at a junction, moves in a regular fashion to form a short path, and ends at a junction.
Short Sequence Digital Video Data
The data that contains the digital video of a short sequence.
A single viewpoint.
Synthetic Viewpoint Image
An image generated as though it were formed from a camera placed at a synthetic viewpoint.
Synthetic viewpoint (also Called a “Novel Viewpoint”)
A camera position (or viewpoint) that is not the position of an actual camera.
A sequence produced from a stationary viewpoint
User Service Process
A computer program or group of programs that serve users.
Same as Viewpoint
A position, describable with 6 parameters (three translational and three rotational) where a camera or eye would be in order to view a scene.
The motion of a viewpoint. Strictly speaking, viewpoints do not move, but are sequentially selected, as though they were moving.
The convincing reality of being somewhere (that one is not) based on the images being displayed that are the same as one would see if one were present in the scene and moving.
Virtual Moving Images
The images displayed during virtual motion.
The present invention is directed to a method for effectively delivering visual telepresence to a plurality of simultaneous users. The method of the present invention involves the efficient processing and delivery of image data in a unique way. More specifically, the processing begins by capturing ongoing electronic images from a plurality of cameras distributed throughout an environment in many fixed locations.
Within the scene, a set of paths is defined, along with junctions, much like a road network. The direct paths between each pair of junctions are called “short paths” and are predetermined. Each short path is then divided into a sequence of viewpoints. If one were to sequentially capture images from the sequence of viewpoints along a short path, they would appear as though one were walking along this short path, for instance, down the sidewalk in a town. Because it is impossible to place cameras at every viewpoint, a smaller number of cameras are placed in fixed positions in the scene, nearby short path junctions. Computer processing techniques use images from these actual cameras to synthesize images as though they were produced from the viewpoints along the short paths.
Much like travelers may traverse long paths through the road network by selecting different roads at junctions, long paths may be constructed through the scene composed of short paths by selecting appropriate short paths at junctions. In the present invention, this process of short path selection takes place without the knowledge of the user. Instead, the user simply decides which direction to proceed, stop, rotate, etc.
A separate computer processing device repetitively produces the sequence of images from synthetic viewpoints along the short path, and many such devices are simultaneously doing so in synchrony.
By connecting the predetermined short path sequences together, a continuous virtual path through the scene is provided to each user or viewer. Each viewer is served by a computer process that composes the digital video stream appropriate to that user's ongoing view. The viewer may decide to change virtual motion, and therefore would require a change in short path sequence. A means is provided in the present invention that routes the digital video stream from each ongoing short sequence virtual view device to the process that serves that viewer. This routing is performed by a routing or switching device as is common in computer networking. Other computer devices and processes oversee the routing so as to balance the load among the user (viewer) processors and the delivery of the correct short path digital video the correct process.
Referring now to the figures, wherein like reference numerals indicate like elements, in
For example, a plurality of electronic imaging devices, such as video or digital cameras (hereinafter referred to as “cameras”), are positioned so that a moving or stationary scene is captured. Two or more images of the scene are of appropriate resolution and view for a desired application. Preferably, each camera is capable of providing a continuous stream of images, as in a video.
Although there is no specific requirement to orient the cameras in a precise manner, there is a minimum requirement that the cameras are oriented such that at least pairs of cameras have view angles that are substantially similar and provide desired viewing coverage.
It is also possible for the camera arrangement to be less carefully placed or randomly placed, where sufficient quantity of overlapping views with sufficient parallax exists to produce short sequences that are joined to form useful paths. Random or nearly random placements of the cameras may be necessary for military or law enforcement applications, or where rapid placement, stealth or covert camera placement is advantageous.
Cameras must be oriented so that at least views from pairs of cameras overlap each other. For instance, as depicted in
Accordingly, camera 8 is positioned to have a view angle 11, which overlaps in coverage with respect to the view of camera 7 with view angle 10. Furthermore, the pair of cameras 7,8 have views that substantially contain the viewpoint sequence 14. In a substantially similar way, camera 9 is positioned to have a view angle 12, which overlaps in coverage with camera 8 and with respect to other short sequences of viewpoints (not depicted) captured by camera 8, and so on.
In this respect, the cameras 7, 8, 9 depicted in
The above-mentioned example is provided to assist in the understanding of the present invention and is not meant to impose limitations on the range of applications or usefulness of the invention.
In another embodiment, with respect to
For instance, if the desired application parameters require simulation of person walking at a leisurely pace, then the configuration density of cameras 7, 8, 9 to capture a person walking at a pace of 1.5 meters per second, is to locate one camera every three meters in order to have a two-second short sequence length. Therefore, the camera density objective here is to place one or more cameras at an average of one every three meters, and to set the short sequence length to two seconds. This placement equates average spatial short sequence lengths to average camera densities.
Similar calculations may be applied to determine the number of cameras having different lens magnification, which is required to capture zooming short sequences effectively. In a similar manner, one can determine a configuration of cameras that may point outward at various angles in a manner necessary to make panning effective. In the cases of zooming and panning, very little or no parallax information is needed and therefore, camera sets that are used for these short segments are preferably co-located.
For example, in viewing a street, where the objective is image coverage of sidewalks on both sides of the street, both sidewalks may be imaged for a walking pace at a density of about 1,000 cameras for 250 meters of street length. In this configuration, approximately every three meters (along both sidewalks) a group of 6 cameras is placed. Greater or lesser densities may be used, depending on the amount of visual obstruction in the area, the likelihood that events of interest will occur that attract users, and the availability of permission or ability to secure cameras.
Again, this example is provided for instruction and is not intended to limit the range of applications or usefulness of the invention.
Referring now to
For instance, the camera arrangement depicted in
Referring back to
Preferably, substantially all short sequences of virtual viewpoints are produced simultaneously and in lock step. The length of time for each short sequence is determined by a combination of application requirements, camera placement and processing availability. A discussion of performance tradeoffs is presented below.
Each short sequence consists of a predetermined set of spatial coordinates and virtual camera parameters that may, in general, constitute a set of viewpoints along some trajectory and with some constant or changing view direction, magnification and rotation about the optical axis. This short sequence may, for example, consist of a straight line of equally spaced viewpoints all having parallel optical axes and identical magnification. Such a linear configuration of the viewpoints creates effective motion of constant velocity. Or, in another example, the viewpoints may follow a curve or rotate about a point, essentially providing a panning view. Alternatively, the viewpoints may keep identical optical axis and change magnification, essentially providing a zoom effect.
It is also possible for a short sequence to simply constitute an unmoving sequence of identical viewpoints, which provides a view as though a viewer is stationary. Although any combination of viewpoints is possible for a specific predetermined short sequence, it is most likely that the sequence would simulate some smooth motion as would be appreciated by a viewer.
Furthermore, each short sequence, save the beginning and last in each terminal of a long sequence, has a synthetic viewpoint beginning at or very near the same location as the end of at least one other sequence, and perhaps the ends of more than one other sequence. Such beginnings and endings are called “Junctures”. Additionally, because it is convenient to place clusters of cameras together, it is likely in many applications that several short sequences will participate in each such juncture. Also, it is advantageous that viewpoints at or near Junctures share common view directions and magnifications, etc. so as to allow smooth visual transitions between predetermined short sequences.
In some circumstances, it is advantageous to have some sequences longer than most or longer than others. For simplicity, the sequences in the preferred embodiment are constrained to lengths that are integer multiples of the common short sequence length. By so doing, the junctures at the beginning and end of longer as well as normal short sequences are all processed at the same instant. The constraint that such a longer short sequence places on the user experience is that, when viewing this particular portion of the virtual path, the command to change virtual paths is delayed by some integer multiple of the normal response delay to the user commands.
The mechanics by which short sequence digital video data are produced are shown in
Generally, in operation, at least two cameras 7, 8 produce images that are delivered to one generator 26. Alternatively, more than two camera sources may be connected to a single sequence video generator such as image 25 is delivered to generator 26. Generator 26 provides an output 33 in synchronization with all other such outputs 36 so that all sequences begin at the same instant and end at the same instant. Note that those short sequences that are by design longer than the majority of short sequences do not participate in all begin/end instants.
More specifically, operation of the sequence video generator 26 is the mechanism by which continuing real electronic images from cameras that view adjacent and overlapping portions of the scene are used to produce a short and repetitive view sequence.
Generator 26 comprises several component processes and data. Many types of processing steps may be used to implement synthetic viewpoint image generation. One example is shown herein where a correspondence finder 26 initiates or updates an existing correspondence field 27. This correspondence field, along with camera parameters 30 and a viewpoint from the viewpoint generator 32 are used by a warp field generator 29 to produce a warp field. A warper 32, warps the image from one of the cameras 23 (suitably delayed to be brought into temporal registration with the warp field data). The viewpoint sequencer 32 provides the specific viewpoint needed for the specific instant.
The warper 32 produces a synthetic viewpoint image 33. All of these processes are ongoing and continuing so that there is no significant temporal jitter or ongoing accumulated delay. Other images, such as 25, may be used in the process of determining pixel values for some portions of the image.
Referring now back to
The routing means 20 is capable of multi-casting; that is, sending the same data source to more than one recipient. The routing means 20 is provided with updates as frequently as the short sequence repetition rate, and a small fraction of connections may be changed during each sequence time period.
Referring now to
Mapping is a one-to-many process in that any input, such as data 36, may map to any single or plurality of outputs, as at 39. However, each output, at any instant, is sourced from one input only.
A subsequent process or device, as in 38, provides controlling commands so that the router/switch may function. The controlling commands update mapping at a frequency of once per short sequence time. Furthermore, this update may be limited by the capacity of the data packet router or matrix switch 37 to perform such changes. Generally, such changes in a map will be on the order of no more than ten percent of the connections per short sequence time.
For example, in situations where there are 1,000 sources of input data 36, 1,000 destinations of output 39, and a sequence length of two seconds, then the controller 38 performs the updating function every two seconds with no more than 100 connection changes.
Note that the present invention is not limited to the above example numbers. Greater numbers of data channels and shorter sequences may be applicable, and more powerful routers/switches may be employed to handle larger numbers of data connections and larger percentage and frequencies of map changes than described in the example.
Referring now to
Each processor 40, 44, 45 provides at least one single visual telepresence. Therefore, a plurality of processors such as thousands of user service processors, is capable of supporting thousands of telepresences. When any specific user desires a change of virtual path, that user process is moved to a processor with sufficient capacity that is servicing the requested virtual path.
The user service means also comprises a user service load-balancing processes 47 and 48, which are configured to switch user processes to the appropriate processor, and to direct the routing of data.
By way of operation, each processor 40, 44, 45 receives a stream of digital visual data 39 from the switching means composed of short path sequences. These data may be composed into a continuous path through the scene. The data 39 are assembled into a continuous stream in a process called video streaming 41, and this stream is delivered over the network to the user via a network or the Internet 50.
The user service application process 42 provides interfaces between the user and any application that might provide other services to the user. It also includes features for control of virtual motion. When a user desires a virtual motion to change, the synthetic position processor 43 coordinates with the user position processor 47, over a local network 46. For instance, if the user wishes to turn or stop, controls on the user graphical interface (such as a web browser) may be activated. In response, the synthetic position processor 43 communicates with the user position process 47, and an appropriate user service process 44, 45, etc. wherein the next desired short sequence is being streamed is selected. The specific user application process is then moved to that user service process via commands over the local area network 46.
The user service load-balancing process 48 balances the load across the various user service processors 40, 44, 45 etc. The load balancing process 48 communicates with a user position processor 47 and a switch control process 49. When the user desires a change of virtual motion, the load balancing process determines whether the targeted processor for the specific user process which contains the desired virtual short sequence has the capacity to accept the additional load of that specific user process. If it does not, then another processor is assigned. Such assignment also requires an update of the router commands as indicated by connection 50.
The method of the present invention may be implemented with existing hardware and/or software. Factors that affect efficient performance of the present method include: 1) the number of short sequences, 2) the length of short sequences, 3) the number of cameras, 4) the switch, 5) the number of user processors, and 6) the number of users.
A primary feature and advantage of the present invention is the efficient presentation of visual telepresence with virtual motion to many users. The primary means by which this is accomplished is by providing a large but finite set of predetermined short sequences that are then joined together to fulfill the desire of the user to navigate. This ability to navigate is constrained to only those paths that are predetermined. Therefore, both the quantity and the shortness of short sequences contribute to the flexibility of user navigation.
The most flexible performance of the present method involves having every possible view throughout a space to be captured to be simultaneously synthesized, thereby bringing the short sequence length to one. In such a situation, the user has substantially total navigational control.
As computational systems become more powerful, it is conceivable that such a system may be constructed, even though there may be the need for millions or billions of such viewpoints. Furthermore, more and shorter sequences are likely to require more cameras to be placed in the environment. It is convenient but not necessary to have approximately one camera per short sequence on average.
In terms of computational capabilities, one pair of cameras may be processed to produce one short sequence by one high performance computer with one attached specialized image processor with warping capability, producing images of the order of 320×240 resolution at 15 frames per second, circa 2002. For the 250-meter, two-sidewalk street example, delivery of a virtual telepresence requires approximately 1,000 such processors for image processing.
When image data processing capabilities allow for higher capacities one should consider whether to increase frame rates or resolution rather than to produce more and shorter sequences. This is a matter of the requirements of the application. Some applications may benefit from greater frame rates and resolution, while others may benefit from greater flexibility in navigation, and hence, more and shorter sequences.
Any increase in the number of short sequences being processed also increases the demand on the switch/router. If the number of virtual telepresences increases, then the number of user service processes also increases. Both of these increases, in turn, tend to increase the number of channels of data to be handled by the switch/router.
Regarding the user service process and processors, existing computer systems may provide one channel of video streaming, and also support about ten users. It is possible that with increased computational power, the number of video streams may be increased, or the number of users increased, or both. If the number of video streams is increased, the incoming channels of digital image data must increase accordingly.
The above description and drawings are only illustrative of preferred embodiments that achieve the objects, features and advantages of the present invention, and it is not intended that the present invention be limited thereto. Any modification of the present invention that comes within the spirit and scope thereof is considered to be part of the present invention.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5259040 *||Oct 4, 1991||Nov 2, 1993||David Sarnoff Research Center, Inc.||Method for determining sensor motion and scene structure and image processing system therefor|
|US5649032 *||Nov 14, 1994||Jul 15, 1997||David Sarnoff Research Center, Inc.||System for automatically aligning images to form a mosaic image|
|US5684935 *||Nov 27, 1995||Nov 4, 1997||Hughes Electronics||Rendering and warping image generation system and method|
|US5850352 *||Nov 6, 1995||Dec 15, 1998||The Regents Of The University Of California||Immersive video, including video hypermosaicing to generate from multiple video views of a scene a three-dimensional video mosaic from which diverse virtual video scene images are synthesized, including panoramic, scene interactive and stereoscopic images|
|US6219444||Feb 2, 1998||Apr 17, 2001||Yissum Research Development Corporation Of The Hebrew University Of Jerusalem||Synthesizing virtual two dimensional images of three dimensional space from a collection of real two dimensional images|
|US6353678 *||Jul 13, 2000||Mar 5, 2002||Sarnoff Corporation||Method and apparatus for detecting independent motion in three-dimensional scenes|
|US6522325 *||Oct 15, 1999||Feb 18, 2003||Kewazinga Corp.||Navigable telepresence method and system utilizing an array of cameras|
|US6667748 *||Apr 21, 1999||Dec 23, 2003||Alcatel||Method of displaying image sequences on mobile radio communications equipment co-operating with a subscriber identity module|
|US6711379 *||May 28, 1999||Mar 23, 2004||Kabushiki Kaisha Toshiba||Digital broadcasting system and terminal therefor|
|US6753899 *||Sep 3, 2002||Jun 22, 2004||Audisoft||Method and apparatus for telepresence|
|US7152210 *||Oct 18, 2000||Dec 19, 2006||Koninklijke Philips Electronics N.V.||Device and method of browsing an image collection|
|US7277118 *||Nov 20, 2001||Oct 2, 2007||Fuji Xerox Co., Ltd.||Method and system for compensating for parallax in multiple camera systems|
|US20030231179 *||Mar 14, 2003||Dec 18, 2003||Norihisa Suzuki||Internet system for virtual telepresence|
|US20040096119||Sep 12, 2003||May 20, 2004||Inoe Technologies, Llc||Efficient method for creating a viewpoint from plurality of images|
|US20040119435 *||Feb 17, 2004||Jun 24, 2004||Pinney Steve B.||Mechanical bug|
|US20070026852 *||Oct 3, 2006||Feb 1, 2007||James Logan||Multimedia telephone system|
|1||Avidan, S. et al, "Novel View Synthesis by Cascading Trilinear Tensors," IEEE Transactions on Visualization and Computer Graphics, vol. 4, No. 4, pp. 293-306, Oct.-Dec. 1998.|
|2||Barrett, E.B. et al., "Some Invariant Linear Methods in Photogrammetry and Model-Matching," IEEE, O-8186-2855, pp. 122-128, Mar. 1992.|
|3||Faugeras, O. et al, "What can two images tell us about a third one?", INRIA, Programme 4-Robotique, image et vision, Projet Robotvis, pp. 1-23, Jul. 1993, Sophia Antipolis, France.|
|4||*||Kanade, Takeo et al., Virtualized Reality: Constructing Virtual Worlds from Real Scenes, Jan.-Mar. 1997, IEEE MultiMedia, 34-46.|
|5||Shashua, A. et al., "Trilinearity of Perspective Views and its Associated Tensor," Institute of Computer Science, Jun. 1995, Hebrew University of Jerusalem.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US8527340||Oct 16, 2012||Sep 3, 2013||Kba2, Inc.||Systems and methods for analytic data gathering from image providers at an event or geographic location|
|US9020832||Mar 7, 2012||Apr 28, 2015||KBA2 Inc.||Systems and methods for analytic data gathering from image providers at an event or geographic location|
|US9264474||May 7, 2013||Feb 16, 2016||KBA2 Inc.||System and method of portraying the shifting level of interest in an object or location|
|U.S. Classification||715/720, 348/14.01, 715/719, 715/716, 715/850, 348/14.04, 348/14.06, 715/782, 345/581, 345/427, 345/594|
|International Classification||H04N7/173, G09G5/00, G06T15/20, G06F3/01, H04N5/247, G06F3/00|
|Cooperative Classification||H04N21/6587, H04N5/247, G06T15/20, H04N21/2343, H04N7/17318, H04N21/472, H04N21/4223|
|European Classification||H04N21/4223, H04N21/2343, H04N21/6587, H04N21/472, H04N5/247, G06T15/20, H04N7/173B2|
|Sep 12, 2003||AS||Assignment|
Owner name: INOE TECHNOLOGIES, LLC, MASSACHUSETTS
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:WILLIAMS, THOMAS D.;HOUGH, ALFRED A.;REEL/FRAME:014492/0475
Effective date: 20020910
|Aug 31, 2010||CC||Certificate of correction|
|May 10, 2013||FPAY||Fee payment|
Year of fee payment: 4